Abstract People living in different areas may differ in their health status because the residents differ in characteristics that influence health (e.g., demographic attributes, socioeconomic status, baseline morbidity) and/or because places differ in characteristics that influence the health of residents, such as neighborhood factors (e.g., local wealth, crime, residential stability) or medical service factors (e.g., physician supply, availability of specialized care). Here, we propose to assemble a data set that will allow us, in subsequent work, to examine the impact of place, defined at two different levels of spatial resolution, on cancer course in the elderly. Our first aim is to build a novel, three-level, hierarchical data set about an inception cohort of elderly Medicare beneficiaries diagnosed with cancer in 11 US regions during 1998-2002. There will be 550,000 individual patients drawn from the SEER cancer registry at level I, linked by Census tract of residence to information about their neighborhoods (N=8,790) at level II, and linked by their ZIP code to information about their health service areas (HSAs) (N=304) at level III. Sources of data include the SEER-Medicare data for level I variables, US Census data and other data for level II variables, and Dartmouth Atlas data for level III variables. Individual- level outcome variables include cancer stage at presentation, receipt of recommended anti-cancer therapies, and survival. The data set will also contain information about the hospitals at which individuals received their cancer care. Our second aim is to examine the relationships among variables measured at the two spatial levels in order to describe, for example, whether rich neighborhoods tend to be clustered in HSAs that are well endowed with medical infrastructure. Ultimately, analyses of this data set will allow us to distinguish the role of the individual (level I) from the roles of the two spatially nested geographic areas, i.e., micro neighborhood (level II) and macro health service area (level III) in determining illness course in cancer patients. Other analyses could also ultimately be done in order to evaluate the possible mediating role of the quality of hospitals used by patients in the relationship between place and individual outcomes.